I plan to call inverse_dynamics in mujoco api and the joint coordinate makes me a little confused. In mujoco document (http://www.mujoco.org/book/), it said one of the key distinctions between MuJoCo and gaming engines (such as ODE, Bullet, Havoc, PhysX) is that MuJoCo operates in generalized or joint coordinates, while gaming engines operate in Cartesian coordinates. To begin with, I tried to set qpos and see whether it looks the same as kinematic data. I used the euler angle in a Cartesian joint coordinate, it looks good except some slight foot sliding. I believe it doesn't introduce problem because qpos is calculated using a T pose as base. I wonder whether it will introduce problem when I calculate the angle velocity. Given the euler angles for each joints of two poses, can I do subtraction on a time sequence and set them as angular velocity, qvel? Any coordinate system transformation is needed?